About the author - terrytaylorbonn/auxdrone GitHub Wiki

25.0928 [email protected], linkedin.com/in/terry-taylor-biz, Lab notes (Gdrive), Git

Eric Schmidt: "I want America to win" (me: "Why not both?"). An interesting discussion about Ziptie topics: AI, LLMs, brilliant Chinese AI tech, Drone AI, and the Ukraine war.

From 25.0929 - 25.1029 I will be visiting Taipei, Hangzhou, and Shenzhen. I may have limited access to my mobile number and internet.

1 Professional focus

Building and deploying software & documentation using the latest AI tools

My current project is the "Pilot Project Pilot" demo. A (WIP) complete start-to-finish AI agent test setup. For details see the docx #493b_pilot-project-pilot_GPT-CPLT_.docx. This will continue to be my focus (after my return from my trip) for at least the rest of 2025.

This wiki documents my AI tool experience. For the past 2 years my efforts have focused mainly on:

LLMs running on local Win11/WSL2 or hosted platforms (OpenAI, HF, etc) (deployments).
MERN stacks (deployments).
CNN object recognition deployed on autonomous drones (Jetson Nano / Pi).
- EPIC 1 - Build/fly FPV drone
- EPIC 2 - Build/fly Pixhawk drone
- EPIC 3 - Add AI to Pixhawk drone

I've recently written 3 docs (WIP) that summarize my conceptual understanding of LLMs and CNNs:

#438c_gpt3_decoded_.docx. How GPT-3 (LLM) simulates intelligence using GPU matrix math (video overview).
#438c-2_ALEXNET_.docx. Computational algorithms for AlexNet (CNN) (v1 25.0914).
#438c-3_GPT3_ALEXNET_comparison_.docx. Similarities of the computational algorithms of GPT-3 and ALEXNET (v1 25.0920).

I'm always amazed by the brilliant engineering behind incredible AI tools that run on binary platforms (GPUs) without a single (biological) neuron (AI will not be conscious | Roger Penrose).

2 Links

AI LLM stack deployed demos.

Lab notes (Gdrive), Git.

Public sites: ZiptieAI.com, web test sites, Substack, Wiki, Youtube.

Recent Substack posts (focusing on example demos for wiki section 4 GPT/CPLT demos):

#35 Tiny LLM (local) + RAG + MCP / Pilot project latest (25.0928).
#34 Pilot project pilot / My dream job (25.0925).
#32 AlexNet <> GPT3 / A comparison of the computational pipelines (25.0919).
#31 AlexNet deep dive / Pipeline computation details (25.0918).
#30 A TL;DR intro to my book (WIP) about decoding GPT-3 computational workflows (25.0913).
#30 FFN (feed forward network) concepts // From the book "GPT3 decoded" (WIP) (25.0828).
#24 ChatGPT as a copilot for Jira // An example of the future of SW tool development (25.0813).
#23 Creating DITA-OT docs // From scratch using only Copilot (no Oxygen) (25.0809). Deployed on Render.
#22 DITA-OT + Copilot demo / How I replaced the Oxygen XML editor with Copilot (25.0807).
#19 Cloudflare LLM deployment demo / Is CF a better choice than AWS/GCP?.
#18 LLM fine-tune/deploy demo . smolLMv3 on HuggingFace step-by-step (25.0726).
#17 MCP concepts/demo. The best MPC intro and step by step demo (25.0721).

Previous project: Autonomous AI (object recognition) drones — early test flights were in my kitchen.
Resources: Website, Wiki, Docs, Gdrive, YouTube

3 Languages

Listed in order of proficiency. I have worked for local private sector tech employers (no NGOs) in all of these countries (except Ukraine).